[prev in list] [next in list] [prev in thread] [next in thread] 

List:       aix-l
Subject:    Re: writing patterns to disk via diag
From:       "F. Even" <aixlists () ELITISTS ! ORG>
Date:       2009-03-27 21:54:18
Message-ID: c5bba06c0903271454i37dd7ec5x5a25711fb2277c0c () mail ! gmail ! com
[Download RAW message or body]

On Mon, Mar 9, 2009 at 1:28 AM, F. Even <aixlists@elitists.org> wrote:
> On Fri, Mar 6, 2009 at 8:57 AM, Flavien <flavien-aix@lebarbe.net> wrote:
>>> I'm trying to wipe disks by writing a bunch of patterns to them from
>>> AIX.  How we normally do it is boot via a NIM install w/ a customized
>>> bosinst.data file.  We have stanzas such as this customized:
>>>
>>>     ERASE_ITERATIONS = 7
>>>     ERASE_PATTERNS = 00,ff,a5,5a,00,ff,a5
>>> [...]
>>>
>>> diag -d hdiskX -c -T format -s erase -a write -P 00,ff,a5,5a,00,ff,a5
>>>
>>> This doesn't seem to be actually writing the patterns from what I can
>>> tell though because well...it finishes way to quick.  It runs through
>>> a 70 gig disk in 20 minutes.
>>
>> If I get my maths correctly, that makes about 60MB/s.
>>
>> Givent current disks performance, it looks good to me.
>>
>> I suspect the NMI install does not set the drive in its fastest
>> operation mode, whereas AIX does it. That would explain the
>> difference you're seeing.
>>
>> PS: Note that you might want to ask diag to do it more than once (if
>>    such an option is available in diag), or launch diag several
>>    times.
>
> Still not sure this is about right though.  I would expect pattern
> writes to take longer than a simple format, but that does not seem to
> be the case.  Not sure my command is doing what I want it to do (may
> have to use one of the other methods suggested):
>
> # time diag -d hdisk4 -c -T format -s erase -a write -P 00,ff,a5,5a,00,ff,a5
> FORMAT MEDIA TASK
>
> Device:  hdisk4 in location
> U7879.001.XXXXXXX-P1-T12-L4-L0
>
> The format operation has completed successfully.
>
>
> real    20m4.78s
> user    0m0.05s
> sys     0m0.10s
>
> # time diag -d hdisk4 -c -T format
> FORMAT MEDIA TASK
>
> Device:  hdisk4 in location
> U7879.001.XXXXXXX-P1-T14-L4-L0
>
> The format operation has completed successfully.
>
>
> real    20m6.73s
> user    0m0.05s
> sys     0m0.10s
>
>
> The pattern write run actually finished quicker by a couple of
> seconds.  Seems wrong.
>

FYI - if anyone was curious, I did get an answer from IBM about this.
A command that actually works for one iteration (and then I can script
it up to just restart for the # of iterations necessary).  This does
what I was looking for:

diag -cd hdisk## -T "format -s erase -a write -P 00,ff,a5,5a,00,ff,a5"


Frank
[prev in list] [next in list] [prev in thread] [next in thread]